isabella232 / financial-services-workload-samples Goto Github PK

Financial Services Workload Samples:
====================================

Most FSI applications are deployed in a multi-process (one process per core) environment. So, all the workloads in this sample will execute in Multi-processes/Instances mode. Each process will execute a single thread and inter-process communication and resources sharing between the threads won't exist while working on separate instruments (i.e. Options). In the package the compilation will build AVX512 binaries from source code. The Intel C++ Compiler is needed to compile the sources.

Black Scholes Option Pricing:
============================= 

Black Scholes is a popular mathematical model used in finance for European option valuation. This is a double precision version. This benchmark prices call and put options using the Black Scholes formula. A large set of options are fed via an input array and results are written into call and put arrays. An exp function is used for simulating the Cumulative Normal Distribution (CND). This is a compute-bound, double precision workload and benefits from Turbo. It runs on Physical cores (HT can be turned off).

Binomial Option Pricing:
========================

Binomial Options Pricing is a lattice-based approach (Cox, Ross and Rubenstein method) that uses a discrete-time model of the varying price over time of the underlying financial instrument (a European call option). At every time step, the value of stock “S” can either go up by uS or go down by “vS” (0<v<1<u). The values of “u” and “v” are constant for every time step. This is a compute-bound, double precision workload and benefits from Turbo and SMT. Run it in HT ON mode (runs on both Physical and Logical (HT) cores).

Monte Carlo Option Pricing:
===========================

Monte Carlo European options is a numerical method that uses statistical sampling techniques to approximate solutions to quantitative problems. In Computational Finance, Monte Carlo algorithms are used to calculate the value of an option with multiple sources of uncertainties and random features, such as changing interest rates, stock prices or exchange rates, etc. to evaluate complex instruments, portfolios, and investments. This is a compute-bound, double precision workload and benefits from Turbo and SMT. Run it in HT ON mode (runs on both Physical and Logical (HT) cores).


Example Run:
============

[user@SUT FSI_SAMPEL_WORKLOADS] # ls
BinomialOptions  BlackScholes  MonteCarloEuropeanOptions
[user@SUT FSI_SAMPEL_WORKLOADS]# cd MonteCarloEuropeanOptions
[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]# ls
Makefile  MonteCarloInsideBlockingDP.cpp  MPTest  REDME.txt
[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]# ls MPTest
clean.sh  getresults.sh  runbatch.sh

[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#source /opt/intel/compilers_and_libraries_........../linux/bin/compilervars.sh intel64
[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#export PR="Number OF Cores in the system reported by lscpu"

//Put numerical value of the cores count showed by #lscpu for Binomial and Monte Carlo with HT ON state. For Black Scholes if HT is OFF then PR will same as total cores number reported by #lscpu command otherwise it will be 50% of the of total cores count shows by #lscpu command with HT ON state.

[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#make
[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#ls
Makefile  MonteCarloInsideBlockingDP.avx512  MonteCarloInsideBlockingDP.cpp  MonteCarloInsideBlockingDP.optrpt  MPTest  REDME.txt
[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#cp MonteCarloInsideBlockingDP.avx512 MPTest
[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#cd MPTest
[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#ls MPTEST
clean.sh  getresults.sh  MonteCarloInsideBlockingDP.avx512  runbatch.sh

[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#./runbatch.sh MonteCarloInsideBlockingDP.avx512

// Need to make sure with #top command to confirm that all the instances(processes) started and completed.

[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]# top
top - 14:04:25 up 3 days,  1:26,  3 users,  load average: 4.57, 1.01, 0.47
Tasks: 996 total,  97 running, 899 sleeping,   0 stopped,   0 zombie
%Cpu(s): 99.8 us,  0.2 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem : 19651051+total, 18213664+free,  2937204 used, 11436676 buff/cache
KiB Swap: 65535996 total, 65535996 free,        0 used. 19252812+avail Mem

   PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
303188 root      20   0   34948   3048   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303189 root      20   0   34948   3052   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303190 root      20   0   34948   3052   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303191 root      20   0   34948   3052   1892 R  87.5  0.0   0:04.87 MonteCarloInsid
303192 root      20   0   34948   3052   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303193 root      20   0   34948   3048   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303194 root      20   0   34948   3048   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303195 root      20   0   34948   3052   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303196 root      20   0   34948   3048   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303197 root      20   0   34948   3048   1892 R  87.5  0.0   0:04.88 MonteCarloInsid
303198 root      20   0   34948   3052   1892 R  87.5  0.0   0:04.88 MonteCarloInsid

//Getting results: invoke getresult.sh script passing with result file name. All data will be dumped in the file and main result will be displayed (execution time in sec)

[root@SUT FSI_SAMPEL_WORKLOADS MPTest]# ./getresults.sh result
7.00145

// The time is the average time to take all instances of the workload (MonteCarlo..) to be completed

[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#ls
clean.sh       inst12  inst18  inst23  inst29  inst34  inst4   inst45  inst50  inst56  inst61  inst67  inst72  inst78  inst83  inst89  inst94
getresults.sh  inst13  inst19  inst24  inst3   inst35  inst40  inst46  inst51  inst57  inst62  inst68  inst73  inst79  inst84  inst9   inst95
inst0          inst14  inst2   inst25  inst30  inst36  inst41  inst47  inst52  inst58  inst63  inst69  inst74  inst8   inst85  inst90  MonteCarloInsideBlockingDP.avx512
inst1          inst15  inst20  inst26  inst31  inst37  inst42  inst48  inst53  inst59  inst64  inst7   inst75  inst80  inst86  inst91  result.txt
inst10         inst16  inst21  inst27  inst32  inst38  inst43  inst49  inst54  inst6   inst65  inst70  inst76  inst81  inst87  inst92  runbatch.sh
inst11         inst17  inst22  inst28  inst33  inst39  inst44  inst5   inst55  inst60  inst66  inst71  inst77  inst82  inst88  inst93

[user@SUT FSI_SAMPEL_WORKLOADS MonteCarloEuropeanOptions]#

Calculation of the performances: Number of options*Number of processes/ Average Time( reported on the screen via running getresults.sh. Unit: Options/sec

Monte Carlo European Options Throughput =Total Number of Options/Average Time needs to complete all processes 
= (Number of Options * NLoops(Iteration)*Nprocesses) / Time"	getresult.sh calculates average time to complete all processes to complete


Binomial Options Throughput =Total Number of Options/Average Time needs to complete all processes 
= (Number of Options * NLoops(Iteration)*Nprocesses) / Time"	getresult.sh calculates average time to complete all processes to complete
		
Black Scholes Throughput =Total Number of Options/Average Time needs to complete all processes 
= (Number of Options * Nloops(Iteration)*Nprocesses) / Time"	getresult.sh calculates average time to complete all processes to complete